Well, what would you expect to be printed by

  var_dump($b < $a)

Based on the 'Transcription of standard array comparison' example on the php Comparison Operators page I'd expect var_dump($b < $a) to be true and var_dump($b > $a) to be false as $b['a'] < $a['a']. In the same way that var_dump($a < $b) should be true and var_dump($a > $b) should be false as $a['b'] < $b['b'].

This does make array ordering comparison not commutative, but that's what the documentation say. It also means that in no case are var_dump($a < $b) and var_dump($a > $b) both equal to true.

if (isset($_SERVER['SERVER_ADDR'])) 
{
	if (strpos($_SERVER['SERVER_ADDR'],"192.168.10") >= 0)
	{
		$where="http://qnap/piwik/";
	}
	else
	{
		$where="https://www.someurl/piwik/";
	} 
} else $where="https://someurl/piwik/";

Problem arrises when >= is processed

why do you come with the same broken snippet as in https://bugs.php.net/bug.php?id=75980 on this topic? your code is simply broken because you don't understand how weak-typing works when return value of a boolean false is completly different meaning then int 0

I can only excuse my self.
I thought that a missinterpreting of >= lead o the problem. 
As you clearly stated it's not to use <? as a start tag as i did.
(which is not stated in the germen documentation)

That my posted source has a problem from improper syntax is also a benefit for me ... but it worked for years without someone complaining or leading not to the desired result

The problem here is that the result of the comparison depends on
which array is the "first" array: hence $a < $b and $b < $a.
However, the engine assumes that $a > $b is equivalent to $b <=
$a, which is certainly a false assumption with regard to array
comparisons.  We most certainly could fix this, and although the
results would match the documentation then, they still would not
make much sense.  Besides, the resulting BC break would be an
issue.

It seems to me this issue deserves some discussion on the
internals@ mailing list.

Re-classifying as feature request as per my comment above.